caffeine_lang - v6.2.0

caffeine_lang/linker/ir

Types

DepsValidated

</>

Marker type for IRs with validated dependencies.

pub type DepsValidated

ExternalValueExtraction

</>

Resolved value-extraction spec on an ExternalSignal indicator. path is the dotted path into the source event (e.g. “value” for a Langfuse score’s numeric value); type_ is the resolved type constraint the extracted value must satisfy.

pub type ExternalValueExtraction {
  ExternalValueExtraction(
    path: String,
    type_: types.AcceptedTypes,
  )
}

Constructors

IndicatorSource

</>

An indicator’s source — either an inline query string (the existing form) or an external signal contract that the relay materializes into a metric the vendor codegen then queries.

LiteralQuery carries a query string (e.g. "sum:requests{...}") with possible $$var$$ template variables; this is the pre-6.x shape.

ExternalSignal carries the routing info for a runtime relay (Langfuse today). At codegen time the DD pipeline rewrites it into a LiteralQuery referencing a synthesized metric name (caffeine.<measurement>.<indicator>), and the relay codegen emits a matching entry in signals.json.

pub type IndicatorSource {
  LiteralQuery(query: String)
  ExternalSignal(
    source: String,
    match: dict.Dict(String, value.Value),
    value_extraction: option.Option(ExternalValueExtraction),
  )
}

Constructors

IntermediateRepresentation

</>

Internal representation of a parsed expectation with metadata and values. The phantom type parameter phase tracks pipeline progress:

  • Linked: freshly built by the linker
  • DepsValidated: dependencies have been validated
  • Resolved: indicators have been resolved by semantic analysis
pub type IntermediateRepresentation(phase) {
  IntermediateRepresentation(
    metadata: IntermediateRepresentationMetaData,
    unique_identifier: String,
    values: List(helpers.ValueTuple),
    slo: SloFields,
    vendor: option.Option(vendor.Vendor),
  )
}

Constructors

IntermediateRepresentationMetaData

</>

Metadata associated with an intermediate representation including organization and service identifiers. Fields use newtype wrappers to prevent accidental mixing of identifier kinds.

pub type IntermediateRepresentationMetaData {
  IntermediateRepresentationMetaData(
    friendly_label: identifiers.ExpectationLabel,
    org_name: identifiers.OrgName,
    service_name: identifiers.ServiceName,
    measurement_name: identifiers.MeasurementName,
    team_name: identifiers.TeamName,
    misc: dict.Dict(String, List(String)),
  )
}

Constructors

Linked

</>

Marker type for IRs freshly built by the linker.

pub type Linked

Resolved

</>

Marker type for IRs with resolved indicators.

pub type Resolved

SloFields

</>

Structured SLO artifact fields extracted from raw values.

description is the SLO description text (sourced from ### doc comments preceding the expectation in the source file). Datadog codegen renders it into the description attribute, optionally combined with the runbook link.

pub type SloFields {
  SloFields(
    threshold: Float,
    indicators: dict.Dict(String, IndicatorSource),
    window_in_days: Int,
    evaluation: option.Option(String),
    tags: List(#(String, String)),
    runbook: option.Option(String),
    depends_on: option.Option(
      dict.Dict(dependency.DependencyRelationType, List(String)),
    ),
    description: option.Option(String),
    below_ms: option.Option(Float),
    expectation_type: option.Option(ast.ExpectationType),
  )
}

Constructors 

  • SloFields(
  threshold: Float,
  indicators: dict.Dict(String, IndicatorSource),
  window_in_days: Int,
  evaluation: option.Option(String),
  tags: List(#(String, String)),
  runbook: option.Option(String),
  depends_on: option.Option(
    dict.Dict(dependency.DependencyRelationType, List(String)),
  ),
  description: option.Option(String),
  below_ms: option.Option(Float),
  expectation_type: option.Option(ast.ExpectationType),
)

    Arguments

    below_ms

    Latency threshold in milliseconds, sourced from a Guarantees N% below <duration> clause. Valid only on time_slice-shaped SLOs; codegen errors if present on a metric SLO.

    expectation_type

    Declared SLO type from the measurement header (success_rate or time_slice). None when the measurement uses the legacy untyped header — semantic checks that need the type (E10 alignment, F13 latency monotonicity) skip pairs where either side is None.

✨ Search Document